Uploaded image for project: 'Hive'
  1. Hive
  2. HIVE-27082

AggregateStatsCache.findBestMatch() in Metastore should test the inclusion of default partition name




      This JIRA deals with non-determinisitic behavior of Hive in generating DAGs.

      The non-determinstic behavior of Hive in generating DAGs is due to the logic in AggregateStatsCache.findBestMatch() called from AggregateStatsCache.get(), as well as the disproportionate distribution of Nulls in HIVE_DEFAULT_PARTITION.

      Here is what is happening in the case of the TPC-DS dataset. Let us use web_sales table and ws_web_site_sk column in the 10TB TPC-DS dataset as a running example.

      In the course of running TPC-DS queries, Hive asks MetaStore about the column statistics of 1823 partNames in the web_sales/ws_web_site_sk combination, either without HIVE_DEFAULT_PARTITION or with HIVE_DEFAULT_PARTITION.

      — Without HIVE_DEFAULT_PARTITION, it reports a total of 901180 nulls.
      — With HIVE_DEFAULT_PARTITION, however, it report a total of 1800087 nulls, almost twice as many.

      The first call to MetaStore returns the correct result, but all subsequent requests are likely to return the same result from the cache, irrespective of the inclusion of HIVE_DEFAULT_PARTITION. This is because AggregateStatsCache.findBestMatch() treats HIVE_DEFAULT_PARTITION in the same way as other partNames, and the difference in the size of partNames[] is just 1. The outcome depends on the duration of intervening queries, so everything is now non-deterministic.

      If a wrong value of numNulls is returned, Hive generates a different DAG which make takes much longer than the correct one. The problem is particularly pronounced here because of the huge number of nulls in HIVE_DEFAULT_PARTITION. It is ironic to see that the query optimizer is so efficient that a single wrong guess of numNulls creates a very inefficient DAG.

      Note that this behavior cannot be avoided by setting hive.metastore.aggregate.stats.cache.max.variance to zero because the difference in the number of partNames[] between the argument and the entry in the cache is just 1.

      So, AggregateStatsCache.findBestMatch() should treat HIVE_DEFAULT_PARTITION in a special way, by not returning the result in the cache if there is a difference in the inclusion of partName HIVE_DEFAULT_PARTITION (or should provide the use with an option to activate this feature).


        Issue Links



              seonggon Seonggon Namgung
              glapark Sungwoo Park
              0 Vote for this issue
              1 Start watching this issue



                Time Tracking

                  Original Estimate - Not Specified
                  Not Specified
                  Remaining Estimate - 0h
                  Time Spent - 20m